JP4627711B2 - Motion region detection system, motion region detection method, and motion region detection program - Google Patents

Description

  The present invention relates to a motion region detection system for detecting a motion region in an image.

  A security service is provided to connect monitoring cameras and various sensors to the center via a network, monitor them, and detect abnormalities when they are detected. In order to realize such a security service, a technique for detecting and recognizing a situation / state in real time such as detecting a suspicious movement is particularly required.

  In view of these problems, a technique for detecting a motion region in a video using image information has been disclosed. As these techniques, a method based on image difference is known. A particular problem here is that the luminance and chromaticity signal differences between the moving object and the background change due to factors such as the shooting environment. As such an example, an example of trying to solve a problem when a shadow or the like is included in the detection target region is known. For example, as a moving object extraction device, by considering an image with low appearance frequency such as a shadow that moves with a moving object and a specular reflection image due to a floor surface, the detection target area may be a mirror reflection image due to a shadow or floor surface, or an object It is said that it is possible to cope with accidental deformation or the like (for example, see Patent Document 1).

However, the above-described techniques only consider changes such as shadows and specular reflections caused by the movement of an object, and are input depending on the illumination conditions of the environment in which the image is captured and the characteristics of the camera used for the image capture. There is a drawback that a motion region in a video cannot be detected correctly when the luminance or chromaticity signal difference changes. In recent years, the performance of a relatively inexpensive and easy-to-use camera such as a WEB camera has not been high, and due to these drawbacks, it is very difficult to stably detect a motion region in an actual environment.
Japanese Patent No. 3625910

  Accordingly, it is an object of the present invention to provide a motion region detection technique for detecting a motion region in a video, particularly a technology that is robust against illumination of a shooting environment and camera noise.

Accordingly, in order to solve the above-mentioned problem, the invention according to claim 1 is a system for detecting a motion region in a video, and includes a video input unit for inputting a video and a motion region detection for generating a motion pattern image. and parts, have a, a motion domain output unit for outputting the motion area with respect to the frame image, the motion area detection unit, an image initialization unit for initializing the motion pattern image, the three from the image An image reading unit that reads an image, a pixel value acquisition unit that acquires a pixel value of each pixel position for all pixel positions of each of the three images, and each of the acquired pixel values between the three images An absolute value calculation unit that calculates a difference ratio and calculates an absolute value, an evaluation value determination unit that uses the absolute value as an evaluation value for a corresponding pixel position, and a motion pattern image that uses the evaluation values at all pixel positions as pixel values A motion pattern image generator for generating Pixel values are assigned in the motion pattern image, a norm calculation unit that calculates a norm of the motion pattern, a motion detection determination unit that determines that motion is detected in the image when the norm satisfies a threshold, and A rectangular region extraction unit that extracts a rectangular region in which pixels continuously exist, and a motion region determination unit that determines the region as a motion region when the size of the rectangular region satisfies a threshold value. Features.

Further, in the invention according to claim 2, when the three images are a first image, a second image, and a third image in order of time, the absolute value calculation unit includes the first image and the first image. The absolute value is calculated by taking the ratio of the difference between the pixel values of the third image and the difference between the pixel values of the second image and the third image.

According to a third aspect of the present invention, there is provided a region detection method in a system for detecting a motion region in a video, wherein the video input unit inputs a video, and the motion region detection unit detects a motion pattern image. a motion region detection step to be generated, the motion area output step by the motion domain output unit for outputting the motion area with respect to the frame image, have a, the motion area detection step, the pattern image image initialization unit motion An image initialization step for initialization, an image reading step for the image reading unit to read three images from the video, and a pixel value acquisition unit for each pixel position of each of the three images. An absolute value calculating step for calculating an absolute value by calculating a ratio of a difference between the acquired pixel values between the three images. An evaluation value determining step in which the value determining unit uses the absolute value as an evaluation value of the corresponding pixel position, and a motion pattern image generation in which the motion pattern image generating unit generates a motion pattern image using the evaluation values at all pixel positions as pixel values. A norm calculation step in which a norm calculation unit calculates a norm of the motion pattern; and a motion detection determination step in which a motion detection determination unit determines that a motion is detected in the image when the norm satisfies a threshold value; A rectangular area extracting unit that extracts a rectangular area in which pixels having pixel values assigned continuously exist in the motion pattern image, and a motion area determining unit sets a threshold value for the size of the rectangular area. A moving region determining step for determining the moving region as a moving region when the region is satisfied;

The invention according to claim 4
When the three images are a first image, a second image, and a third image in order of time,
The absolute value calculating step calculates an absolute value by taking a ratio of a difference between pixel values of the first image and the third image and a difference between pixel values of the second image and the third image. And

  The invention according to claim 5 is characterized in that the motion region detection method is described by a computer program and can be executed.

  According to the first to fifth aspects of the present invention, it is possible to detect a motion region in an image, and to provide a technique that is robust against illumination of a shooting environment and camera noise.

  Hereinafter, embodiments will be described with reference to the drawings.

  FIG. 1 is a functional block diagram of the motion region detection system. As illustrated in FIG. 1, the motion region detection system 20 includes a video input unit 21, a motion region detection unit 22, and a motion region output unit 23.

  In the video input unit 21, a video including a captured motion is input and each frame image is stored in a memory. The input image can use a camera 10 using a CCD as an image sensor.

  The motion region output unit 23 outputs a motion region for each frame image stored in the video input unit 21 based on the processing result in the motion region detection unit 22. The motion region is not particularly limited as long as it is information including the motion region in each frame image in the target video, and the motion pattern generated by the motion region detection unit 22 of the motion region detection system 20 is not limited. It can include images and information obtained by normal image processing.

  FIG. 2 is a flowchart showing the operation of the motion region detection unit 22 of the motion region detection system 20, and FIG. 3 is a functional block diagram of the motion region detection unit 22. The motion region detection unit 22 generates a motion pattern image from the three frame images stored in the video input unit 21. The operation will be described below with reference to FIGS.

  [Step 1] The image initialization unit 31 initializes the pixel value of the motion pattern image to zero.

  [Step 2] The frame image reading unit 32 reads frame images I (0), I (1), and I (2) for three frames from the latest one from the memory.

  [Step 3] The pixel position acquisition unit 33 uses the pixel values p (x, y, 0), p (x, y, I) at the pixel positions (x, y) of I (0), I (1), I (2). 1), p (x, y, 2) is acquired.

  [Step 4] The absolute value is calculated by taking the ratio of the differences between the pixel values acquired by the absolute value calculator 34.

  [Step 5] The evaluation value determination unit sets ΔP (x, y) as the evaluation value of the pixel position (x, y).

  [Step 6] Perform [Step 3] to [Step 5] for all pixel positions of the acquired frame image.

  [Step 7] The motion pattern image generation unit 36 generates a motion pattern image using the evaluation values at all pixel positions as pixel values.

  [Step 8] The norm calculation unit 37 calculates the norm of the motion pattern image.

  [Step 9] Motion Detection Determining Unit 38 When the calculated norm satisfies the threshold, I (0) is determined to have detected motion.

  [Step 10] When a movement is detected by the rectangular area extraction unit 39, a rectangular area in which pixels assigned pixel values are continuously present in the movement pattern image is extracted.

  [Step 11] When the size of the rectangular area satisfies the threshold value, the motion area determining unit 40 determines that it is a motion area.

  At this time, the threshold used in [Step 9] and [Step 11] may be determined in advance, or dynamically using a statistically calculated value such as an average value, a frequency, or a deviation value. It may be calculated as follows. Further, the process of extracting the rectangular area is the same as the area division process used in normal image processing, and these methods are not particularly different from the conventional methods, and thus detailed description thereof is omitted here. .

  Below, the specific procedure at the time of detecting the motion area in an image | video in the motion area detection system 20 demonstrated so far is demonstrated.

  Here, a specific processing procedure when the motion region detection unit 22 generates a motion pattern image will be described. Here, for the sake of explanation, it is assumed that the image read from the memory is a 3 frame image of 5 × 5 pixel size as indicated by 401 to 403 in FIG. 4. 401 corresponds to I (0), 402 corresponds to I (1), and 403 corresponds to I (2). At this time, the pixel values are different even at the same position other than the area of the moving object due to the influence of the illumination of the photographing environment and the noise of the camera.

  For each pixel, a pixel value at each pixel position is acquired and a difference is obtained. For the sake of explanation, the results are represented by images as indicated by 404 and 405. At this time, 404 corresponds to I (0) and I (2), and 405 corresponds to I (1) and I (2). When a ratio of these difference values is taken to calculate an absolute value and a motion pattern image having the absolute value as a pixel value is generated, an image as shown by 406 is obtained.

  A norm is calculated for this motion pattern image. Here, it is determined that a motion has been detected, assuming that the threshold is satisfied. Subsequently, when a rectangular region in which pixels assigned pixel values are continuously present is extracted, a region indicated by hatching in 501 in FIG. 5 is extracted. Here, assuming that the size of this rectangular area satisfies the threshold value, it is determined as a motion area.

  In the above-described embodiment, the motion region detection system is realized by, for example, a CPU included in a computer device that configures the motion region detection system, and a motion region detection process or the like can be installed as an application program.

  In addition, processing results and calculation results such as pixel values, absolute values, evaluation values, motion image patterns, norms, motion detection, rectangular region extraction, and motion region determination performed in motion region detection processing are stored in the main storage device. You may do it.

  In addition, data such as pixel values, absolute values, evaluation values, motion image patterns, norms, motion detection, rectangular region extraction, motion region determination, and information necessary for calculation can be written to and read from an auxiliary storage device or the like. You may do it.

  In addition, a recording medium in which a program code of software for realizing the function of the motion region detection system is recorded is supplied to the system or apparatus, and the CPU (MPU) of the system or apparatus stores the program code stored in the storage medium. It is also possible to execute reading. In this case, the program code itself read from the storage medium realizes the functions of the above-described embodiment, and examples of the storage medium storing the program code include CD-ROM, DVD-ROM, and CD-R. , CD-RW, MO, and HDD.

  The motion region detection system has been described above, but is not limited to the above-described embodiment, and can be widely applied.

  As described above, according to the motion region detection system, a motion region in an image can be detected, and a system that is robust against illumination of a shooting environment and camera noise can be provided.

The functional block diagram of a motion area detection system. The processing flowchart which shows operation | movement of a motion area | region detection part. The functional block diagram of a motion area | region detection part. Explanatory drawing of an example of a process of a motion area | region detection system. Explanatory drawing of an example of a process of a motion area | region detection system.

Explanation of symbols

DESCRIPTION OF SYMBOLS 10 Camera 20 Motion area detection system 21 Image | video input part 22 Motion area detection part 23 Motion area output part 31 Image initialization part 32 Frame image reading part 33 Pixel value acquisition part 34 Absolute value calculation part 35 Evaluation value determination part 36 Motion pattern image Generation unit 37 Norm calculation unit 38 Motion detection unit 39 Rectangular region extraction unit 40 Motion region determination unit

Claims (5)

A system for detecting a motion area in an image,
A video input section for inputting video;
A motion region detection unit for generating a motion pattern image;
Have a, a motion domain output unit for outputting the motion area with respect to the frame image,
The motion region detection unit
An image initialization unit for initializing the motion pattern image;
An image reading unit for reading three images from the video;
A pixel value acquisition unit that acquires a pixel value of each pixel position for all pixel positions of each of the three images;
An absolute value calculation unit that calculates the absolute value by taking a ratio of the difference between the acquired pixel values between the three images;
An evaluation value determination unit that uses the absolute value as an evaluation value of the corresponding pixel position;
A motion pattern image generation unit that generates a motion pattern image using the evaluation values of all pixel positions as pixel values;
A norm calculation unit for calculating a norm of the motion pattern;
A motion detection determining unit that determines that motion is detected in the image when the norm satisfies a threshold;
In the motion pattern image, a rectangular area extraction unit that extracts a rectangular area in which pixels assigned pixel values are continuously present;
A motion region detection system comprising: a motion region determination unit that determines a region as a motion region when the size of the rectangular region satisfies a threshold value.   When the three images are a first image, a second image, and a third image in order of time,
  The absolute value calculation unit calculates an absolute value by taking a ratio of a difference between pixel values of the first image and the third image and a difference between pixel values of the second image and the third image. The motion region detection system according to claim 1. An area detection method in a system for detecting a motion area in a video,
A video input step in which the video input unit inputs video;
A motion region detection step in which the motion region detection unit generates a motion pattern image;
Possess a motion area output step of moving area output unit outputs the motion area with respect to the frame image, and
The moving region detecting step includes
An image initialization step in which the image initialization unit initializes the motion pattern image; and
An image reading step in which the image reading unit reads three images from the video;
A pixel value acquisition step in which the pixel value acquisition unit acquires the pixel value of each pixel position for all pixel positions of each of the three images;
An absolute value calculating step in which an absolute value calculating unit calculates the absolute value by taking a ratio of the difference between the acquired pixel values between the three images;
An evaluation value determining step in which the evaluation value determining unit sets the absolute value as an evaluation value of the corresponding pixel position;
A motion pattern image generation step in which the motion pattern image generation unit generates a motion pattern image having the evaluation values of all pixel positions as pixel values;
A norm calculating step in which a norm calculating unit calculates a norm of the motion pattern;
A motion detection determining step in which a motion detection determining unit determines that motion is detected in the image when the norm satisfies a threshold;
A rectangular area extracting step in which a rectangular area extracting unit extracts a rectangular area in which pixels assigned pixel values are continuously present in the motion pattern image;
A motion region detection method comprising: a motion region determination step for determining, when a size of the rectangular region satisfies a threshold, a motion region determination unit, when the motion region determination unit satisfies the threshold value.   When the three images are a first image, a second image, and a third image in order of time,
  The absolute value calculating step calculates an absolute value by taking a ratio of a difference between pixel values of the first image and the third image and a difference between pixel values of the second image and the third image. The motion region detection method according to claim 3.   5. A motion region detection program according to claim 3, wherein the motion region detection method according to claim 3 is described in a computer program so as to be executable.